Three-Dimensional (3-D) Segmentation of Knee Cartilage from 3.0 T MR Images using a Graph-Cuts Technique

Introduction Osteoarthritis (OA), which is a common cause of disability, is associated with degradation of articular cartilage. Quantitative evaluation of knee cartilage requires segmentation of the cartilage which is challenging and laborious. For instance, manual delineation of cartilage boundaries by an expert often results in several hours per case and is subject to analyst bias and error. Hence, we have developed a semi-automated approach based on a graph-cuts algorithm [1] for efficient and reproducible segmentation of knee cartilage from high-resolution 3T MR images. We also evaluated the performance of our semi-automated graph-cuts method and compared it to the performance of the conventional manual delineation segmentation method.